Autenticación de canela verdadera (Cinnamomum verum) usando imágenes hiperespectrales y quimiometría
No Thumbnail Available
Date
2024
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad Nacional de Trujillo
Abstract
La canela, una preciada especia con fragantes propiedades, desempeña un papel fundamental tanto en la industria alimentaria como en la farmacéutica. En el mercado, es posible encontrar dos variantes principales de canela: innamomum verum, conocida como canela verdadera, exclusiva de Sri Lanka, y Cinnamomum cassia, llamada canela falsa, que se cultiva en diversas regiones geográficas. Por ende, el propósito de este estudio consistió en desarrollar modelos de clasificación basados en imágenes hiperespectrales NIR (infrarrojo cercano) acopladas a técnicas quimiométricas. El objetivo era distinguir entre ramas de canela C. verum y C. cassia. Para empezar, se aplicó el análisis de componentes principales (PCA) para explorar las imágenes hiperespectrales. Los resultados del análisis de scores indicaron una alta similitud entre las especies, respaldada por concentraciones similares de macronutrientes. El tercer componente principal (PC3) permitió una diferenciación más efectiva de las clases en comparación con PC1 y PC2. Este componente presentó cargas que revelaron picos asociados a compuestos fenólicos y aromáticos, como la cumarina en el caso de C. cassia y la catequina en el caso de C. verum. Mediante el uso del análisis discriminante por mínimos cuadrados parciales (PLS-DA), se logró un buen desempeño en la clasificación según el origen de las muestras, con un índice de error del 3,3 % y una precisión del 96,7 %. Una prueba de permutación con un valor p < 0,05 confirmó que las predicciones de PLS-DA se basan en las características reales de los datos espectrales, descartando la posibilidad de resultados casuales. Los mapas de clasificación, tanto por píxeles (enfoque A) como por muestras (enfoque B, C y D), alcanzaron una tasa de clasificación correcta del 98,3 % para C. verum y del 100 % para C. cassia. Estos resultados indican que la combinación de NIR-HSI y PLS-DA se erige como un enfoque analítico confiable para llevar a cabo la autenticación de la canela.
ABSTRACT Cinnamon, a prized spice with fragrant properties, plays a crucial role in both the food and pharmaceutical industries. In the market, two main variants of cinnamon can be found: Cinnamomum verum, known as true cinnamon, exclusive to Sri Lanka, and Cinnamomum cassia, referred to as cassia cinnamon, cultivated in various geographical regions. Therefore, the purpose of this study was to develop classification models based on near-infrared hyperspectral (NIR-HSI) images coupled with chemometric techniques. The goal was to distinguish between branches of C. verum and C. cassia cinnamon. To commence, principal component analysis (PCA) was applied to explore the hyperspectral images. The score analysis results indicated a high similarity between the species, supported by similar concentrations of macronutrients. The third principal component (PC3) allowed for a more effective class differentiation compared to PC1 and PC2. This component revealed loadings that unveiled peaks associated with phenolic and aromatic compounds, such as coumarin in the case of C. cassia and catechin in the case of C. verum. By employing partial least squares discriminant analysis (PLS-DA), a strong performance was achieved in classifying the samples' origin, with an error rate of 3.3% and an accuracy of 96.7%. A permutation test with a p-value of < 0.05 confirmed that PLS-DA predictions are grounded in the actual spectral data characteristics, eliminating the possibility of random outcomes. The classification maps, both by pixels (Approach A) and by samples (Approaches B, C, and D), attained a correct classification rate of 98.3% for C. verum and 100% for C. cassia. These results indicate that the combination of NIR-HSI and PLS-DA emerges as a reliable analytical approach for carrying out cinnamon authentication.
ABSTRACT Cinnamon, a prized spice with fragrant properties, plays a crucial role in both the food and pharmaceutical industries. In the market, two main variants of cinnamon can be found: Cinnamomum verum, known as true cinnamon, exclusive to Sri Lanka, and Cinnamomum cassia, referred to as cassia cinnamon, cultivated in various geographical regions. Therefore, the purpose of this study was to develop classification models based on near-infrared hyperspectral (NIR-HSI) images coupled with chemometric techniques. The goal was to distinguish between branches of C. verum and C. cassia cinnamon. To commence, principal component analysis (PCA) was applied to explore the hyperspectral images. The score analysis results indicated a high similarity between the species, supported by similar concentrations of macronutrients. The third principal component (PC3) allowed for a more effective class differentiation compared to PC1 and PC2. This component revealed loadings that unveiled peaks associated with phenolic and aromatic compounds, such as coumarin in the case of C. cassia and catechin in the case of C. verum. By employing partial least squares discriminant analysis (PLS-DA), a strong performance was achieved in classifying the samples' origin, with an error rate of 3.3% and an accuracy of 96.7%. A permutation test with a p-value of < 0.05 confirmed that PLS-DA predictions are grounded in the actual spectral data characteristics, eliminating the possibility of random outcomes. The classification maps, both by pixels (Approach A) and by samples (Approaches B, C, and D), attained a correct classification rate of 98.3% for C. verum and 100% for C. cassia. These results indicate that the combination of NIR-HSI and PLS-DA emerges as a reliable analytical approach for carrying out cinnamon authentication.
Description
Keywords
Especias, Fraude alimentario, Quimiometría, Prueba de permutación, Autenticación